Prior predictive simulation

Elizabeth King

Kevin Middleton

Goals

Do my priors make sense at all?
Are my priors too restrictive?
Are my priors too loose?

Is the simulated data generally on the same scale as the outcome?

Simulate data

You know the answer ahead
Get the answer you know is there

Useful for working through new techniques

Simulate DBH for 100 trees

Log Normal distribution
- log mean = 2.5 (12.2 cm mean)
- log sd = 0.5 (1.6 cm standard deviation)

set.seed(34729)

DBH <- rlnorm(100, 2.5, 0.5)
DBH

  [1] 15.052024 10.516033 48.768134  6.051757 23.568834  8.215073  8.358373
  [8] 10.965030 18.721478  6.185453 33.463990  7.692346  8.183408 38.299289
 [15] 19.662011 40.735673 47.761948  8.941139 14.348609 17.649784 10.298022
 [22] 16.744447 18.485528  4.401309 20.393710  9.387616 18.634552  3.833338
 [29] 12.991330 14.447092 10.030150 34.974844  5.855171 13.695729  6.122422
 [36]  7.711018 15.457268 12.589072 23.771346 13.454850 16.433005  7.669530
 [43]  7.016928 10.147020 10.392751 14.755238  5.863227 11.664974 37.834532
 [50]  8.122134 11.637684  8.054722 10.671032 11.067064 19.764847 11.350978
 [57]  5.528933 11.334868 10.751546  7.264224 11.535853  7.576272 38.627318
 [64]  8.357703 17.193855 12.844281 14.871730 23.751698 19.244272 15.178204
 [71] 25.143548  9.448907  5.198409 18.509202  6.271502  4.401013  8.829788
 [78]  3.829214 20.244310  7.092822 15.633127  7.137421 15.143549  8.846810
 [85] 11.810036 19.420051 14.109783 12.172618 20.902973 16.982590 14.497679
 [92]  8.214659 23.667628 11.506670 10.022089  8.371849 11.861825  6.422960
 [99] 28.380817 12.254432

Visualize

ggplot(tibble(DBH), aes(DBH)) +
  geom_histogram(bins = 30)

Visualize

log-transform?

ggplot(tibble(log_DBH = log(DBH)), aes(log_DBH)) +
  geom_histogram(bins = 30)

log-transform?

Model statements 1

\[DBH \sim Normal(\mu, \sigma)\]

No additional predictors
Mean only
Priors for \(\mu\) and \(\sigma\)

Prior prediction

Generate many simulated datasets

\(DBH \sim Normal(mean, sd)\)
mean and sd are drawn from distributions themselves

Plot the distributions of the datasets
Decide if the priors are appropriate

Simulating datasets #1

Half-normal is a common distribution for standard deviations
Use truncnorm::rtruncnorm() to generate a half-normal distribution
a = 0 sets the lower bound at 0

set.seed(3247922)
Sim_data <- purrr::map(
  .x = 1:20,
  .f = function(ii) {
    tibble(ID = ii,
           DBH = rnorm(100,
                       mean = rnorm(1, 10, 2.5),
                       sd = rtruncnorm(5, mean = 2, sd = 1, a = 0)))
  }) |> 
  list_rbind()
Sim_data

Simulating datasets #1

# A tibble: 2,000 × 2
      ID   DBH
   <int> <dbl>
 1     1  8.37
 2     1  7.83
 3     1 13.0 
 4     1 10.7 
 5     1 18.4 
 6     1 10.3 
 7     1 10.4 
 8     1 12.1 
 9     1 11.4 
10     1 13.4 
# … with 1,990 more rows

Half-normal distribution

Plotting #1

ggplot(tibble(DBH), aes(DBH)) +
  geom_histogram(aes(y = after_stat(density)), fill = "#2D5DA1", bins = 30) +
  geom_line(data = Sim_data, aes(DBH, group = ID),
            stat = "density", alpha = 0.5, linewidth = 1)

Plotting #1

Simulating datasets #2

Shift the mean to the right
Increase spread

set.seed(87349)
Sim_data <- purrr::map(
  .x = 1:20,
  .f = function(ii) {
    tibble(ID = ii,
           DBH = rnorm(100,
                       mean = rnorm(1, 12, 5),
                       sd = rtruncnorm(5, mean = 5, sd = 3, a = 0)))
  }) |> 
  list_rbind()

Plotting #2

Relationships between distributions

Simulate with LogNormal #1

Start with small values for prior standard deviations

set.seed(4623864)
Sim_data <- purrr::map(
  .x = 1:20,
  .f = function(ii) {
    tibble(ID = ii,
           DBH = rlnorm(100,
                        mean = rnorm(1, 1, 0.01),
                        sd = rtruncnorm(1, mean = 1, sd = 0.1, a = 0)))
  }) |> 
  list_rbind()

Plotting #3

Simulate with LogNormal #2

Shift the mean right
Spread the standard deviation

set.seed(623469)
Sim_data <- purrr::map(
  .x = 1:20,
  .f = function(ii) {
    tibble(ID = ii,
           DBH = rlnorm(100,
                        mean = rnorm(1, 3, 0.2),
                        sd = rtruncnorm(1, mean = 1, sd = 0.25, a = 0)))
  }) |> 
  list_rbind()

Plotting #4

Model statements

\[DBH \sim LogNormal(\mu, \sigma)\]

\[\mu \sim Normal(3, 0.2)\] \[\sigma \sim HalfNormal(1, 0.25)\]

Define and sample the model

rethinking::ulam() converts model statements to stan code and samples the model.

fm <- ulam(
  alist(
    DBH ~ lognormal(mu, sigma),
    mu ~ normal(3, 0.2),
    sigma ~ normal(1, 0.25)
  ),
  data = list(DBH = DBH),
  chains = 4, cores = 4,
  iter = 5e3,
  seed = 4353479
)

Define and sample the model

Running MCMC with 4 parallel chains, with 1 thread(s) per chain...

Chain 1 Iteration:    1 / 5000 [  0%]  (Warmup) 
Chain 1 Iteration:  100 / 5000 [  2%]  (Warmup) 
Chain 1 Iteration:  200 / 5000 [  4%]  (Warmup) 
Chain 1 Iteration:  300 / 5000 [  6%]  (Warmup) 
Chain 1 Iteration:  400 / 5000 [  8%]  (Warmup) 
Chain 1 Iteration:  500 / 5000 [ 10%]  (Warmup) 
Chain 1 Iteration:  600 / 5000 [ 12%]  (Warmup) 
Chain 1 Iteration:  700 / 5000 [ 14%]  (Warmup) 
Chain 1 Iteration:  800 / 5000 [ 16%]  (Warmup) 
Chain 1 Iteration:  900 / 5000 [ 18%]  (Warmup) 
Chain 1 Iteration: 1000 / 5000 [ 20%]  (Warmup) 
Chain 1 Iteration: 1100 / 5000 [ 22%]  (Warmup) 
Chain 1 Iteration: 1200 / 5000 [ 24%]  (Warmup) 
Chain 1 Iteration: 1300 / 5000 [ 26%]  (Warmup) 
Chain 1 Iteration: 1400 / 5000 [ 28%]  (Warmup) 
Chain 1 Iteration: 1500 / 5000 [ 30%]  (Warmup) 
Chain 1 Iteration: 1600 / 5000 [ 32%]  (Warmup) 
Chain 1 Iteration: 1700 / 5000 [ 34%]  (Warmup) 
Chain 1 Iteration: 1800 / 5000 [ 36%]  (Warmup) 
Chain 1 Iteration: 1900 / 5000 [ 38%]  (Warmup) 
Chain 1 Iteration: 2000 / 5000 [ 40%]  (Warmup) 
Chain 1 Iteration: 2100 / 5000 [ 42%]  (Warmup) 
Chain 1 Iteration: 2200 / 5000 [ 44%]  (Warmup) 
Chain 1 Iteration: 2300 / 5000 [ 46%]  (Warmup) 
Chain 1 Iteration: 2400 / 5000 [ 48%]  (Warmup) 
Chain 1 Iteration: 2500 / 5000 [ 50%]  (Warmup) 
Chain 1 Iteration: 2501 / 5000 [ 50%]  (Sampling) 
Chain 1 Iteration: 2600 / 5000 [ 52%]  (Sampling) 
Chain 1 Iteration: 2700 / 5000 [ 54%]  (Sampling) 
Chain 1 Iteration: 2800 / 5000 [ 56%]  (Sampling) 
Chain 1 Iteration: 2900 / 5000 [ 58%]  (Sampling) 
Chain 1 Iteration: 3000 / 5000 [ 60%]  (Sampling) 
Chain 1 Iteration: 3100 / 5000 [ 62%]  (Sampling) 
Chain 1 Iteration: 3200 / 5000 [ 64%]  (Sampling) 
Chain 1 Iteration: 3300 / 5000 [ 66%]  (Sampling) 
Chain 1 Iteration: 3400 / 5000 [ 68%]  (Sampling) 
Chain 1 Iteration: 3500 / 5000 [ 70%]  (Sampling) 
Chain 1 Iteration: 3600 / 5000 [ 72%]  (Sampling) 
Chain 1 Iteration: 3700 / 5000 [ 74%]  (Sampling) 
Chain 1 Iteration: 3800 / 5000 [ 76%]  (Sampling) 
Chain 1 Iteration: 3900 / 5000 [ 78%]  (Sampling) 
Chain 1 Iteration: 4000 / 5000 [ 80%]  (Sampling) 
Chain 1 Iteration: 4100 / 5000 [ 82%]  (Sampling) 
Chain 1 Iteration: 4200 / 5000 [ 84%]  (Sampling) 
Chain 1 Iteration: 4300 / 5000 [ 86%]  (Sampling) 
Chain 1 Iteration: 4400 / 5000 [ 88%]  (Sampling) 
Chain 1 Iteration: 4500 / 5000 [ 90%]  (Sampling) 
Chain 1 Iteration: 4600 / 5000 [ 92%]  (Sampling) 
Chain 1 Iteration: 4700 / 5000 [ 94%]  (Sampling) 
Chain 1 Iteration: 4800 / 5000 [ 96%]  (Sampling) 
Chain 1 Iteration: 4900 / 5000 [ 98%]  (Sampling) 
Chain 1 Iteration: 5000 / 5000 [100%]  (Sampling) 
Chain 2 Iteration:    1 / 5000 [  0%]  (Warmup) 
Chain 2 Iteration:  100 / 5000 [  2%]  (Warmup) 
Chain 2 Iteration:  200 / 5000 [  4%]  (Warmup) 
Chain 2 Iteration:  300 / 5000 [  6%]  (Warmup) 
Chain 2 Iteration:  400 / 5000 [  8%]  (Warmup) 
Chain 2 Iteration:  500 / 5000 [ 10%]  (Warmup) 
Chain 2 Iteration:  600 / 5000 [ 12%]  (Warmup) 
Chain 2 Iteration:  700 / 5000 [ 14%]  (Warmup) 
Chain 2 Iteration:  800 / 5000 [ 16%]  (Warmup) 
Chain 2 Iteration:  900 / 5000 [ 18%]  (Warmup) 
Chain 2 Iteration: 1000 / 5000 [ 20%]  (Warmup) 
Chain 2 Iteration: 1100 / 5000 [ 22%]  (Warmup) 
Chain 2 Iteration: 1200 / 5000 [ 24%]  (Warmup) 
Chain 2 Iteration: 1300 / 5000 [ 26%]  (Warmup) 
Chain 2 Iteration: 1400 / 5000 [ 28%]  (Warmup) 
Chain 2 Iteration: 1500 / 5000 [ 30%]  (Warmup) 
Chain 2 Iteration: 1600 / 5000 [ 32%]  (Warmup) 
Chain 2 Iteration: 1700 / 5000 [ 34%]  (Warmup) 
Chain 2 Iteration: 1800 / 5000 [ 36%]  (Warmup) 
Chain 2 Iteration: 1900 / 5000 [ 38%]  (Warmup) 
Chain 2 Iteration: 2000 / 5000 [ 40%]  (Warmup) 
Chain 2 Iteration: 2100 / 5000 [ 42%]  (Warmup) 
Chain 2 Iteration: 2200 / 5000 [ 44%]  (Warmup) 
Chain 2 Iteration: 2300 / 5000 [ 46%]  (Warmup) 
Chain 2 Iteration: 2400 / 5000 [ 48%]  (Warmup) 
Chain 2 Iteration: 2500 / 5000 [ 50%]  (Warmup) 
Chain 2 Iteration: 2501 / 5000 [ 50%]  (Sampling) 
Chain 2 Iteration: 2600 / 5000 [ 52%]  (Sampling) 
Chain 2 Iteration: 2700 / 5000 [ 54%]  (Sampling) 
Chain 2 Iteration: 2800 / 5000 [ 56%]  (Sampling) 
Chain 2 Iteration: 2900 / 5000 [ 58%]  (Sampling) 
Chain 2 Iteration: 3000 / 5000 [ 60%]  (Sampling) 
Chain 2 Iteration: 3100 / 5000 [ 62%]  (Sampling) 
Chain 2 Iteration: 3200 / 5000 [ 64%]  (Sampling) 
Chain 2 Iteration: 3300 / 5000 [ 66%]  (Sampling) 
Chain 2 Iteration: 3400 / 5000 [ 68%]  (Sampling) 
Chain 2 Iteration: 3500 / 5000 [ 70%]  (Sampling) 
Chain 2 Iteration: 3600 / 5000 [ 72%]  (Sampling) 
Chain 2 Iteration: 3700 / 5000 [ 74%]  (Sampling) 
Chain 2 Iteration: 3800 / 5000 [ 76%]  (Sampling) 
Chain 2 Iteration: 3900 / 5000 [ 78%]  (Sampling) 
Chain 2 Iteration: 4000 / 5000 [ 80%]  (Sampling) 
Chain 2 Iteration: 4100 / 5000 [ 82%]  (Sampling) 
Chain 2 Iteration: 4200 / 5000 [ 84%]  (Sampling) 
Chain 2 Iteration: 4300 / 5000 [ 86%]  (Sampling) 
Chain 2 Iteration: 4400 / 5000 [ 88%]  (Sampling) 
Chain 2 Iteration: 4500 / 5000 [ 90%]  (Sampling) 
Chain 2 Iteration: 4600 / 5000 [ 92%]  (Sampling) 
Chain 2 Iteration: 4700 / 5000 [ 94%]  (Sampling) 
Chain 2 Iteration: 4800 / 5000 [ 96%]  (Sampling) 
Chain 2 Iteration: 4900 / 5000 [ 98%]  (Sampling) 
Chain 2 Iteration: 5000 / 5000 [100%]  (Sampling) 
Chain 3 Iteration:    1 / 5000 [  0%]  (Warmup) 
Chain 3 Iteration:  100 / 5000 [  2%]  (Warmup) 
Chain 3 Iteration:  200 / 5000 [  4%]  (Warmup) 
Chain 3 Iteration:  300 / 5000 [  6%]  (Warmup) 
Chain 3 Iteration:  400 / 5000 [  8%]  (Warmup) 
Chain 3 Iteration:  500 / 5000 [ 10%]  (Warmup) 
Chain 3 Iteration:  600 / 5000 [ 12%]  (Warmup) 
Chain 3 Iteration:  700 / 5000 [ 14%]  (Warmup) 
Chain 3 Iteration:  800 / 5000 [ 16%]  (Warmup) 
Chain 3 Iteration:  900 / 5000 [ 18%]  (Warmup) 
Chain 3 Iteration: 1000 / 5000 [ 20%]  (Warmup) 
Chain 3 Iteration: 1100 / 5000 [ 22%]  (Warmup) 
Chain 3 Iteration: 1200 / 5000 [ 24%]  (Warmup) 
Chain 3 Iteration: 1300 / 5000 [ 26%]  (Warmup) 
Chain 3 Iteration: 1400 / 5000 [ 28%]  (Warmup) 
Chain 3 Iteration: 1500 / 5000 [ 30%]  (Warmup) 
Chain 3 Iteration: 1600 / 5000 [ 32%]  (Warmup) 
Chain 3 Iteration: 1700 / 5000 [ 34%]  (Warmup) 
Chain 3 Iteration: 1800 / 5000 [ 36%]  (Warmup) 
Chain 3 Iteration: 1900 / 5000 [ 38%]  (Warmup) 
Chain 3 Iteration: 2000 / 5000 [ 40%]  (Warmup) 
Chain 3 Iteration: 2100 / 5000 [ 42%]  (Warmup) 
Chain 3 Iteration: 2200 / 5000 [ 44%]  (Warmup) 
Chain 3 Iteration: 2300 / 5000 [ 46%]  (Warmup) 
Chain 3 Iteration: 2400 / 5000 [ 48%]  (Warmup) 
Chain 3 Iteration: 2500 / 5000 [ 50%]  (Warmup) 
Chain 3 Iteration: 2501 / 5000 [ 50%]  (Sampling) 
Chain 3 Iteration: 2600 / 5000 [ 52%]  (Sampling) 
Chain 3 Iteration: 2700 / 5000 [ 54%]  (Sampling) 
Chain 3 Iteration: 2800 / 5000 [ 56%]  (Sampling) 
Chain 3 Iteration: 2900 / 5000 [ 58%]  (Sampling) 
Chain 3 Iteration: 3000 / 5000 [ 60%]  (Sampling) 
Chain 3 Iteration: 3100 / 5000 [ 62%]  (Sampling) 
Chain 3 Iteration: 3200 / 5000 [ 64%]  (Sampling) 
Chain 3 Iteration: 3300 / 5000 [ 66%]  (Sampling) 
Chain 3 Iteration: 3400 / 5000 [ 68%]  (Sampling) 
Chain 3 Iteration: 3500 / 5000 [ 70%]  (Sampling) 
Chain 3 Iteration: 3600 / 5000 [ 72%]  (Sampling) 
Chain 3 Iteration: 3700 / 5000 [ 74%]  (Sampling) 
Chain 3 Iteration: 3800 / 5000 [ 76%]  (Sampling) 
Chain 3 Iteration: 3900 / 5000 [ 78%]  (Sampling) 
Chain 3 Iteration: 4000 / 5000 [ 80%]  (Sampling) 
Chain 3 Iteration: 4100 / 5000 [ 82%]  (Sampling) 
Chain 3 Iteration: 4200 / 5000 [ 84%]  (Sampling) 
Chain 3 Iteration: 4300 / 5000 [ 86%]  (Sampling) 
Chain 3 Iteration: 4400 / 5000 [ 88%]  (Sampling) 
Chain 3 Iteration: 4500 / 5000 [ 90%]  (Sampling) 
Chain 3 Iteration: 4600 / 5000 [ 92%]  (Sampling) 
Chain 3 Iteration: 4700 / 5000 [ 94%]  (Sampling) 
Chain 3 Iteration: 4800 / 5000 [ 96%]  (Sampling) 
Chain 3 Iteration: 4900 / 5000 [ 98%]  (Sampling) 
Chain 3 Iteration: 5000 / 5000 [100%]  (Sampling) 
Chain 4 Iteration:    1 / 5000 [  0%]  (Warmup) 
Chain 4 Iteration:  100 / 5000 [  2%]  (Warmup) 
Chain 4 Iteration:  200 / 5000 [  4%]  (Warmup) 
Chain 4 Iteration:  300 / 5000 [  6%]  (Warmup) 
Chain 4 Iteration:  400 / 5000 [  8%]  (Warmup) 
Chain 4 Iteration:  500 / 5000 [ 10%]  (Warmup) 
Chain 4 Iteration:  600 / 5000 [ 12%]  (Warmup) 
Chain 4 Iteration:  700 / 5000 [ 14%]  (Warmup) 
Chain 4 Iteration:  800 / 5000 [ 16%]  (Warmup) 
Chain 4 Iteration:  900 / 5000 [ 18%]  (Warmup) 
Chain 4 Iteration: 1000 / 5000 [ 20%]  (Warmup) 
Chain 4 Iteration: 1100 / 5000 [ 22%]  (Warmup) 
Chain 4 Iteration: 1200 / 5000 [ 24%]  (Warmup) 
Chain 4 Iteration: 1300 / 5000 [ 26%]  (Warmup) 
Chain 4 Iteration: 1400 / 5000 [ 28%]  (Warmup) 
Chain 4 Iteration: 1500 / 5000 [ 30%]  (Warmup) 
Chain 4 Iteration: 1600 / 5000 [ 32%]  (Warmup) 
Chain 4 Iteration: 1700 / 5000 [ 34%]  (Warmup) 
Chain 4 Iteration: 1800 / 5000 [ 36%]  (Warmup) 
Chain 4 Iteration: 1900 / 5000 [ 38%]  (Warmup) 
Chain 4 Iteration: 2000 / 5000 [ 40%]  (Warmup) 
Chain 4 Iteration: 2100 / 5000 [ 42%]  (Warmup) 
Chain 4 Iteration: 2200 / 5000 [ 44%]  (Warmup) 
Chain 4 Iteration: 2300 / 5000 [ 46%]  (Warmup) 
Chain 4 Iteration: 2400 / 5000 [ 48%]  (Warmup) 
Chain 4 Iteration: 2500 / 5000 [ 50%]  (Warmup) 
Chain 4 Iteration: 2501 / 5000 [ 50%]  (Sampling) 
Chain 4 Iteration: 2600 / 5000 [ 52%]  (Sampling) 
Chain 4 Iteration: 2700 / 5000 [ 54%]  (Sampling) 
Chain 4 Iteration: 2800 / 5000 [ 56%]  (Sampling) 
Chain 4 Iteration: 2900 / 5000 [ 58%]  (Sampling) 
Chain 4 Iteration: 3000 / 5000 [ 60%]  (Sampling) 
Chain 4 Iteration: 3100 / 5000 [ 62%]  (Sampling) 
Chain 4 Iteration: 3200 / 5000 [ 64%]  (Sampling) 
Chain 4 Iteration: 3300 / 5000 [ 66%]  (Sampling) 
Chain 4 Iteration: 3400 / 5000 [ 68%]  (Sampling) 
Chain 4 Iteration: 3500 / 5000 [ 70%]  (Sampling) 
Chain 4 Iteration: 3600 / 5000 [ 72%]  (Sampling) 
Chain 4 Iteration: 3700 / 5000 [ 74%]  (Sampling) 
Chain 4 Iteration: 3800 / 5000 [ 76%]  (Sampling) 
Chain 4 Iteration: 3900 / 5000 [ 78%]  (Sampling) 
Chain 4 Iteration: 4000 / 5000 [ 80%]  (Sampling) 
Chain 4 Iteration: 4100 / 5000 [ 82%]  (Sampling) 
Chain 4 Iteration: 4200 / 5000 [ 84%]  (Sampling) 
Chain 4 Iteration: 4300 / 5000 [ 86%]  (Sampling) 
Chain 4 Iteration: 4400 / 5000 [ 88%]  (Sampling) 
Chain 4 Iteration: 4500 / 5000 [ 90%]  (Sampling) 
Chain 4 Iteration: 4600 / 5000 [ 92%]  (Sampling) 
Chain 4 Iteration: 4700 / 5000 [ 94%]  (Sampling) 
Chain 4 Iteration: 4800 / 5000 [ 96%]  (Sampling) 
Chain 4 Iteration: 4900 / 5000 [ 98%]  (Sampling) 
Chain 4 Iteration: 5000 / 5000 [100%]  (Sampling) 
Chain 1 finished in 0.1 seconds.
Chain 2 finished in 0.1 seconds.
Chain 3 finished in 0.1 seconds.
Chain 4 finished in 0.1 seconds.

All 4 chains finished successfully.
Mean chain execution time: 0.1 seconds.
Total execution time: 0.3 seconds.

Summarizing the model

summary(fm)

Inference for Stan model: ulam_cmdstanr_2c9c44adfe28e9c6b44eeb19d995e263-202302200926-1-595662.
4 chains, each with iter=5000; warmup=2500; thin=1; 
post-warmup draws per chain=2500, total post-warmup draws=10000.

        mean se_mean   sd   2.5%    25%    50%    75%  97.5% n_eff Rhat
mu      2.55    0.00 0.06   2.45   2.52   2.55   2.59   2.66  6030    1
sigma   0.58    0.00 0.04   0.50   0.55   0.58   0.61   0.67  5432    1
lp__  -88.75    0.02 1.01 -91.50 -89.13 -88.44 -88.04 -87.77  3712    1

Samples were drawn using NUTS(diag_e) at Mon Feb 20 09:26:53 2023.
For each parameter, n_eff is a crude measure of effective sample size,
and Rhat is the potential scale reduction factor on split chains (at 
convergence, Rhat=1).

Median of the LogNormal is ~2.5
Median standard deviation is ~0.5

Posterior predictive check

We have 10,000 posterior distributions.

Prior distributions

Prior prediction

If this seems like a lot of fuss
- You get more experienced
- More critical for more complex models
- We will learn tools to help with the process

Couldn’t we just calculate the mean and standard deviation of the data and use that?
- Short answer: No.

Get to know your data very well.